Electroencephalogram-Based Satisfaction Assessment Brain-Computer Interface in Emerging Video Service by Using Graph Representation Learning.

Abstract

Background: Emerging video services (EVS) offer users various multimedia presentations, and satisfaction assessment is crucial for enhancing their user experience and competitiveness. However, existing research methods are unable to provide a quantitative satisfaction assessment. Electroencephalogram (EEG), as a popular signal source in brain-computer interface (BCI), with the advantage of being difficult to disguise and containing rich brain activity information, has gained increasing attention from researchers. This article aims to investigate the advantages of employing EEG for modeling satisfaction in EVS. Unlike the subjective metrics assessment in traditional video services, generating satisfaction in EVS involves a range of cognitive functions, including cognitive load, emotion, and audiovisual perception, which are difficult to characterize using a single feature. The representation of brain states for complex cognitive functions has been a major challenge for EEG modeling approaches. Methods: To address this challenge, we propose an EEG-based EVS satisfaction assessment BCI by raising a Point-to-Global graph representation learning strategy (P2G) that efficiently identifies satisfaction level through a parallel coding module and a graph-based brain region perception module. P2G captures satisfaction-sensitive graph representations in EEG samples based on coding and integrating point features and the global topography. Results: We validate the effectiveness of introducing a P2G learning strategy in EVS satisfaction modeling using a self-constructed dataset and a relevant public dataset, and our method outperforms existing methods. Additionally, we provide a detailed visual analysis to unveil neural markers associated with EVS satisfaction, thereby laying a scientific foundation for the optimization and development of video services.